Signal Processing for Cognitive Radios

SUDHARMAN K. JAYAWEERA earned his BE in Electrical Engineering from the University of Melbourne, Australia. He earned his MA and PhD degrees in Electrical Engineering from Princeton University, USA. He is currently an Associate Professor in Electrical Engineering at the University of New Mexico, Albuquerque, NM, USA. His research expertise is in signal processing and wireless communications. DR. JAYAWEERA is a senior member of the IEEE. Currently he serves as the Associate Editor of IEEE Transactions on Vehicular Technology.

• Preface xvPart I Introduction to Cognitive Radios 11 Introduction 31.1 Introduction 31.2 Signal Processing and Cognitive Radios 41.3 Software-Defined Radios 61.3.1 Software-Defined Radio Platforms 141.3.2 Software-Defined Radio Systems 151.4 From Software-Defined Radios to Cognitive Radios 191.4.1 The Spectrum Scarcity Problem 191.4.2 Emergence of CRs 211.5 What this Book is About 221.6 Summary 262 The Cognitive Radio 272.1 Introduction 272.2 A Functional Model of a Cognitive Radio 302.2.1 Spectrum Knowledge Acquisition (Spectrum Awareness) 302.2.2 Communications Decision-Making 332.2.3 Learning in Cognitive Radios 332.3 The Cognitive Radio Architecture 352.3.1 Spectrum Sensing Region of a Cognitive Engine 362.3.2 Radio Reconfiguration Region of a Cognitive Engine 362.3.3 Learning Region of a Cognitive Engine 372.3.4 Memory Region of a Cognitive Engine 372.4 The Ideal Cognitive Radio 382.5 Signal Processing Challenges in Cognitive Radios 392.6 Summary 403 Cognitive Radios and Dynamic Spectrum Sharing 423.1 Introduction 423.2 Interference and Spectrum Opportunities 463.3 Dynamic Spectrum Access 503.4 Dynamic Spectrum Leasing 543.5 Challenges in DSS Cognitive Radios 553.6 Cognitive Radios and Future of Wireless Communications 603.7 Summary 61Part II theoretical foundations 654 Introduction to Detection Theory 674.1 Introduction 674.2 Optimality Criteria: Bayesian versus Non-Bayesian 714.2.1 The Bayesian Approach 724.2.2 A Non-Bayesian Approach: Neyman–Pearson Optimality Criterion 734.3 Parametric Signal Detection Theory 754.3.1 Bayesian Optimal Detection 764.3.2 Neyman–Pearson Optimal Detection 824.3.3 Another Non-Bayesian Alternative: The Generalized Likelihood Ratio Test 994.3.4 Parametric Signal Detection in Additive Noise 1034.4 Nonparametric Signal Detection Theory 1224.4.1 Signal Detection in Additive Zero-Median Noise: The Sign Test 1244.4.2 Signal Detection in Additive Symmetric Noise: The Rank Test 1254.4.3 Signal Detection in Additive Zero Median, Zero Mean, Finite-Variance Noise: The t-Test 1264.5 Summary 1275 Introduction to Estimation Theory 1325.1 Introduction 1325.2 Random Parameter Estimation: Bayesian Estimation 1345.2.1 Minimum Mean-Squared Error Estimation 1345.2.2 MMSE Estimation of Vector Parameters 1355.2.3 Linear Minimum Mean-Squared Error Estimation 1385.2.4 Maximum A Posteriori Probability Estimation 1395.3 Nonrandom Parameter Estimation 1405.3.1 Theory of Minimum Variance Unbiased Estimation 1425.3.2 Best Linear Unbiased Estimator 1475.3.3 Maximum Likelihood Estimation 1525.3.4 Performance Bounds: Cramer-Rao Lower Bound 1545.4 Summary 1586 Power Spectrum Estimation 1646.1 Introduction 1646.2 PSD Estimation of a Stationary Discrete-Time Signal 1686.2.1 Correlogram Method 1686.2.2 Periodogram Method 1706.2.3 Performance of the Periodogram PSD Estimate 1726.3 Blackman–Tukey Estimator of the Power Spectrum 1776.4 Other PSD Estimators Based on Modified Periodograms 1816.4.1 Bartlett PSD Estimator 1816.4.2 Welch PSD Estimator 1836.5 PSD Estimation of Nonstationary Discrete-Time Signals 1866.5.1 Temporally Windowed Observations 1886.5.2 Temporal and Spectral Smoothing of PSD Estimates of Nonstationary Discrete-Time Signals 1896.5.3 DFT-Based PSD Computation 1916.6 Spectral Correlation of Cyclostationary Signals 1926.6.1 Spectral Correlation and Spectral Autocoherence 1966.6.2 Time-Averaged Spectral Correlation 1976.6.3 Estimation of Spectral Correlation 1986.7 Summary 2007 Markov Decision Processes 2077.1 Introduction 2077.2 Markov Decission Processes 2097.3 Finite-Horizon MDPs 2127.3.1 Definitions 2127.3.2 Optimal Policies for MDPs 2167.4 Infinite-Horizon MDPs 2227.4.1 Stationary Optimal Policies for Infinite-Horizon MDPs 2247.4.2 Bellman-Optimality Equations 2277.5 Partially Observable Markov Decision Processes 2327.5.1 Definitions 2337.5.2 Policy Evaluation for a Finite-Horizon POMDP 2387.5.3 Optimality Equations for a Finite-Horizon POMDP 2417.5.4 Optimal Policy Computation for a Finite-Horizon POMDP 2427.5.5 Infinite-Horizon POMDPs 2577.6 Summary 2598 Bayesian Nonparametric Classification 2698.1 Introduction 2698.2 K-Means Classification Algorithm 2748.3 X-Means Classification Algorithm 2768.4 Dirichlet Process Mixture Model 2788.4.1 Dirichlet Process 2788.4.2 Construction of the Dirichlet Process 2798.4.3 DPMM 2828.5 Bayesian Nonparametric Classification Based on the DPMM and the Gibbs Sampling 2838.5.1 DPMM-Based Classification of Scalar Observations 2878.5.2 DPMM-Based Classification of Multidimensional Gaussian Observations 2988.5.3 DPMM-Based Classification of Possibly Non-Gaussian Multidimensional Observations 3088.6 Summary 315Part III signal processing in cognitive radios 3219 Wideband Spectrum Sensing 3239.1 Introduction 3239.2 Wideband Spectrum Sensing Problem 3259.3 Wideband Spectrum Scanning Problem 3269.4 Spectrum Segmentation and Subbanding 3289.5 Wideband Spectrum Sensing Receiver 3309.5.1 Homodyne Receiver Configuration 3329.5.2 Super Heterodyne Digital Receiver Configuration 3349.5.3 A/D Conversion and the Discrete-Time Received Signal Model 3359.6 Subband Selection Problem in Wideband Spectrum Sensing 3369.6.1 Subband Dynamics 3389.6.2 A POMDP Model for Subband Selection 3409.6.3 An Optimal Subband Selection Policy for Spectrum Sensing 3479.6.4 A Reduced-Complexity Optimal Sensing Decision-Making Algorithm with Independent Channels 3509.6.5 A Reduced Complexity Optimal Sensing Decision-Making Algorithm with Independent Subbands 3549.6.6 Optimal Myopic Sensing Decision Policies 3549.7 A Reduced Complexity Optimal Subband Selection Framework with an Alternative Reward Function 3559.7.1 A New Model for Subband Dynamics 3579.7.2 A Simplified Reward Function and a Reduced-Complexity Optimal Policy 3599.7.3 A Reduced Complexity Optimal Policy for Independent Subbands 3629.7.4 Optimal Myopic Policies with Reduced Dimensional Subband State Vectors 3639.8 Machine-Learning Aided Subband Selection Policies 3649.8.1 Q-Learning 3659.8.2 Q-Learning in a POMDP: A Q-Learning Algorithm for Subband Selection 3689.9 Summary 37210 Spectral Activity Detection in Wideband Cognitive Radios 37710.1 Introduction 37710.2 Optimal Wideband Spectral Activity Detection 37910.3 Wideband Spectral Activity Detection 38610.4 Wavelet Transform-Based Wideband Spectral Activity Detection 39210.4.1 Wavelet Transform 39410.4.2 Edge Detection with Wavelet Transform 39510.4.3 Spectral Activity Detection Based on Edge Detection 39710.5 Wideband Spectral Activity Detection in Non-Gaussian Noise 39810.5.1 Arbitrary but Known Noise Distribution 39910.5.2 Robust Spectral Activity Detection 40610.6 Wideband Spectral Activity Detection with Compressive Sampling 41310.6.1 Compressive Sampling 41510.6.2 Compressive Sensing of Wideband Spectrum 41910.7 Summary 42111 Signal Classification in Wideband Cognitive Radios 42911.1 Introduction 42911.2 Signal Classification Problem in a Wideband Cognitive Radio 43111.3 Feature Extraction for Signal Classification 43511.3.1 Carrier/Center Frequency 43511.3.2 Cyclostationary Features 43611.3.3 Modulation Type and Order Features 44111.4 A Signal Classification Architecture for a Wideband Cognitive Radio 44511.5 Bayesian Nonparametric Signal Classification 44711.6 Sequential Bayesian Nonparametric Signal Classification 46211.7 Summary 46912 Primary Signal Detection in DSA Cognitive Networks 47212.1 Introduction 47212.2 Spectrum Sensing Problem in Dynamic Spectrum Sharing CR Networks 47512.3 Autonomous Spectrum Sensing for Dynamic Spectrum Sharing 47912.3.1 Secondary User Sensing Observations 48012.3.2 Channel-State (Idle/Busy) Decisions 48112.4 Limitations of Autonomous Spectrum Sensing 48912.5 Cooperative Spectrum Sensing for Dynamic Spectrum Sharing 49212.6 Cooperative Channel-State Detection 49512.6.1 Local Processing and Sensing Reports from Secondary Users 49812.6.2 Final Channel-State Decisions at the SSDC: Decision Fusion 50212.7 Summary 51613 Spectrum Decision-Making in DSA Cognitive Networks 51913.1 Introduction 51913.2 Primary Channel Dynamic Model 52013.3 Sensing Decisions in DSS Networks with Autonomous Cognitive Radios 52213.3.1 Optimal Sensing Policy Determination 52513.3.2 Optimal Myopic Sensing Policy Determination 53013.4 Sensing Decisions in Cooperative DSS Networks 53313.4.1 Optimal SSDC Decisions for Independent Channel Dynamics 53713.4.2 Optimal Myopic Sensing Decisions at the SSDC with Independent Channel Dynamics 54113.5 Summary 55014 Dynamic Spectrum Leasing in Cognitive Radio Networks 55314.1 Introduction 55314.2 DSL with Direct Rewards to Primary Users 55514.2.1 Interference at the Primary Receiver 56014.2.2 A Game Model for Dynamic Spectrum Leasing 56514.2.3 Nash Equilibria in Noncooperative Games 57014.2.4 Existence of a Nash Equilibrium in the DSL Game 57314.3 DSL Based on Asymmetric Cooperation with Primary Users 58714.3.1 A Primary–Secondary Coexistence Model 58814.3.2 Asymmetric Cooperative Communications-Based DSL between Primary Users and a Centralized Secondary Network 59114.3.3 Asymmetric Cooperative Communications-Based DSL between Primary Users and Autonomous Cognitive Secondary Users 60414.4 Summary 60915 Cooperative Cognitive Communications 61315.1 Introduction 61315.2 Cooperative Spectrum Sensing 61915.3 Cooperative Spectrum Sensing and Channel-Access Decisions 62115.4 Cooperative Communications Strategies in Cognitive Radio Networks 62415.5 Asymmetric Cooperative Relaying in DSA Cognitive Radios 62715.5.1 Secondary User Optimal Power Allocation for Asymmetric Cooperative Relaying 62915.5.2 Centralized Assignment of Cognitive Radios for Cooperation with Primary Users: An Ideal Approach 63515.5.3 Centralized Assignment of Cognitive Radios for Cooperation with Primary Users: A Realistic Approach 64015.6 Summary 64416 Machine Learning in Cognitive Radios 64716.1 Introduction 64716.2 Artificial Neural Networks 65016.2.1 Learning Algorithms for LTUs 65116.2.2 Layered Neural Networks 65516.2.3 Learning in Layered Feed-Forward Networks: Back-Propagation Algorithm 65616.2.4 Neural Networks in Cognitive Radios 66216.3 Support Vector Machines 66416.3.1 Statistical Learning Theory 66516.3.2 Structural Risk Minimization with Support Vector Machines 66916.3.3 Linear Support Vector Machines 67016.3.4 Nonlinear Support Vector Machines 67416.3.5 Kernel Function Implementation of Support Vector Machines 67716.3.6 SVMs in Cognitive Radios 67916.4 Reinforcement Learning 68116.4.1 Temporal Difference Learning 68316.4.2 Q-Learning in a POMDP: Replicated Q-Learning 68416.4.3 Reinforcement Learning in Cognitive Radios 68616.5 Multiagent Learning 68816.5.1 Game-Theoretic Multiagent Learning 69116.5.2 Cooperative Multiagent Learning 69416.5.3 Multiagent Learning in Cognitive Radio Networks 69616.6 Summary 698Appendix A Nyquist Sampling Theorem 704Appendix B A Collection of Useful Probability Distributions 711B.1 Univariate Distributions 711B.2 Multivariate Distributions 713Appendix C Conjugate Priors 716References 721Index 740